Table 2. Comparative performance of PMut2017 predictor.
| Method | Coverage (%) | Accuracy | Specificity | Sensitivity | AUC | MCC |
|---|---|---|---|---|---|---|
| SIFT (18) | 89.6 | 0.61 | 0.33 | 0.88 | 0.60 | 0.25 |
| Polyphen2 (19) | 92.1 | 0.64 | 0.35 | 0.91 | 0.63 | 0.32 |
| PROVEAN (20) | 91.5 | 0.64 | 0.41 | 0.87 | 0.64 | 0.31 |
| FATHMM (24) | 90.5 | 0.55 | 0.45 | 0.64 | 0.55 | 0.09 |
| PON-P2 (25) | 42.4 | 0.72 | 0.52 | 0.9 | 0.71 | 0.45 |
| CADD (29) | 95.0 | 0.65 | 0.33 | 0.94 | 0.64 | 0.35 |
| M-CAP (36) | 91.5 | 0.60 | 0.19 | 0.95 | 0.57 | 0.22 |
| Condel (37) | 91.0 | 0.63 | 0.40 | 0.84 | 0.62 | 0.26 |
| LRT (21)a | 95.1 | 0.73 | 0.58 | 0.87 | 0.73 | 0.47 |
| MutationAssessor (22) a | 95.1 | 0.63 | 0.46 | 0.78 | 0.62 | 0.26 |
| MetaSVM (34) a | 95.1 | 0.63 | 0.51 | 0.74 | 0.62 | 0.26 |
| MetaLR (34) a | 95.1 | 0.6 | 0.46 | 0.73 | 0.60 | 0.20 |
| MutationTaster (35) a | 95.1 | 0.65 | 0.31 | 0.96 | 0.64 | 0.36 |
| PMut | 100.0 | 0.71 | 0.65 | 0.76 | 0.71 | 0.42 |
| PMut (85%)b | 81.0 | 0.76 | 0.76 | 0.77 | 0.76 | 0.53 |
| PMut (90%)b | 51.2 | 0.81 | 0.78 | 0.84 | 0.81 | 0.62 |
| PMut (ClinVar)c | 100.0 | 0.73 | 0.88 | 0.85 | 0.75 | 0.49 |
Blind validation based on new variants added to SwissVar during 2016 (3166 variants), CADD predictor has been evaluated using a threshold of 20. AUC: area under the ROC curve, MCC: Matthews correlation coefficient.
aAnalysis performed from ANNOVAR data (42).
bAnalysis restricted to most reliable PMut predictions (reliability level in parentheses).
cBlind validation based on variants reported on ClinVar (43), not present in the SwissVar dataset (20,308 variants). Indicated coverage is calculated on ClinVar dataset.